Page 157 - 南京医科大学自然版
P. 157

第45卷第12期            严舒桐,刘 琪. 机器学习在类风湿性关节炎诊疗及并发症预测中的研究进展[J].
                 2025年12月                    南京医科大学学报(自然科学版),2025,45(12):1834-1844                      ·1839 ·


                性,后续则需要纳入更多数据并进行多中心验证。                            研究团队开发了一款临床组织病理学成像评估基
                    即使基于ML诊疗RA仍存在诸多风险与挑战,                         础模型,已在全球多家医院验证了其对肿瘤标本的
                但其依然具有一定的意义和价值,持续优化 ML 搭                          诊断准确性,预计在2~3年内推出商业化工具                   [70] ;另
                建诊疗RA的可推广、可持续性模型具有提高RA患                           外北京邮电大学科研团队通过多种医学数据和综
                者早期诊断率以及改善预后的潜能。如哈佛大学                             合评估微调构建的大型语言模型 ClinicalGPT,在训
                                                   表2 ML赋能诊断RA及其并发症
                                        Table 2 ML empowers diagnosis of RA and its complications

                      Algorithm                                Modeling indicator                       Reference
                ML diagnosis of RA
                  RF                Ultrasound scoring system ⁃US18                                      [12]
                                    Clinical baseline data
                  ResNet50          Clinical and ultrasound scoring system ⁃EULAR⁃OMERACT                [13]
                                    Clinical blood flow within the synovium(0-3)
                                    Synovial tissue hyperplasia and hypertrophy(0-3)
                  SVM,LASSO         Clinical baseline data                                               [15]
                                    QCT pulmonary scan
                  RF,SVM            Clinical baseline data                                               [17]
                                    Joint characteristics:joint deformity,tender joint count,swollen joint count,DAS28 score,
                                                   joint function,and joint radiological staging
                                    laboratory indicators:circulating anti⁃citrullinated peptide antibodies,etc.
                                    Treatment situation:methotrexate,hormone therapy,etc.
                  DCNN              Chest X⁃ray image                                                    [18]
                                                                             +
                  SVM               Combined detection of RA characteristic genes for CD4 T cells:LOC731186,CR748316, [20]
                                    LDHA,IGFL2,CMAH,MUC1,PDCD1,PIM1,SOCS3,SBNO2,BCL3,and NOG
                  WGCNA,LASSO,      Combined detection of RA characteristic genes for CD8 T cells:GDF15,IGLC1,IGHM,[23-24]
                                                                             +
                  SVM⁃RFE RF        CD8A,GZMA,and PRF1
                  XGBoost           Combined detection of RA characteristic genes for T cells:MIER1,PPP1CB,ICOS, [27]
                                    GADD45A,CD3D,SLFN5,PIP4K2A,and IL6ST
                  LASSO             Combined detection of RA characteristic genes for platelet:MAPK3,ACTB,ACTG1,VAV2, [25]
                                    PTPN6,and ACTN1
                  LASSO             m6A methylation regulatory factors:IGF2BP3 and YTHDC2                [29]
                  SVM,KNN,RF,Logit Serum exosomes LncRNAs:DLEU2,FAM13A⁃AS1,MEG3,and SNHG15               [30]
                  GLMVQ             Synovial chemokines:CXCL4 and CXCL7                                  [26]
                  WGCNA,LASSO,      Synovial genes:RRM2,DLGAP5,KIF11,AKR1C3,MCEE,POLE4,and PFKM          [33]
                  SVM⁃RFE RF
                  RF,LASSO,PLS      Clinical baseline data                                               [34]
                                    Blood tests:blood cell count,erythrocyte sedimentation rate,C⁃reactive protein,immunoglobu⁃
                                    lin,lactate dehydrogenase,hydroxybutyrate dehydrogenase,KL⁃6,D⁃dimer,fibrinogen,fibrin
                                    degradation products,etc.
                                    Tumor markers include carbohydrate antigens(CA19⁃9,CA242)and carcinoembryonic antigen
                                   (CEA)
                  LASSO,SVM         Genetic testing for depression in patients with RA:AURKA,BTN3A2,CXCL10,ERAP2, [37]
                                    MARCO,PLA2G7,EAF1,SDCBP,and RNF19B
                  LASSO,RF,LR       Laboratory tests:blood cell count,number of immune cells,lipoproteins,cholesterol,etc.  [41]
                  RF,ENR            Genetic testing for rheumatoid arthritis with atherosclerosis:let⁃7c⁃5p,miR⁃30e⁃5p,miR⁃4446⁃  [43]
                                    3p,miR⁃126⁃5p,miR⁃3168,miR⁃425⁃5p,miR⁃126⁃3p,miR⁃30a⁃5p and miR⁃125a⁃5p
                Monitoring the efficacy
                of drugs using ML
                  RF,LASSO          Serum exosomes:has⁃circ0002715 and has⁃circ0001946                   [52]
                  LSTM              Determine the patient visit rate:meteorological factors,air pollutants,and patient historical  [57]
                                    visit data
   152   153   154   155   156   157   158   159   160   161   162